State Space Truncation: How Algorithmic Control Collapses the Behavioural Repertoire in Organisational Networks

Algorithmic management has fundamentally reorganized the structural architecture of the modern enterprise, promising unprecedented operational efficiency through the data-driven optimization of human and non-human capital. Algorithmic management has been shown, however, to be a violation of the fundamental cybernetic principle of requisite variety. The optimization represented by algorithmic management constitutes an artificial restriction of the organizational state space. The present paper argues that the use of top-down algorithmic control limits the behavioural repertoire of autonomous agents in a network and therefore violates the principle of requisite variety. In order to reduce variance and operational slack and to increase historical efficiency, algorithmic management creates an environment of extreme epistemological closure, making the network extremely vulnerable to out of distribution environmental shocks. A computational multi agent model of a complex adaptive system was created to test this hypothesis. The system modelled the organization as being located in a stochastic and high-entropy environment. Two different network architectures were modelled; one where the organization’s agents had a limited behavioural repertoire due to a top-down algorithmic constraint on their behaviour and another where the agents’ behaviour was unconstrained allowing them to create a wide range of behaviours. The simulated agents used the principles of active inference and continually attempted to minimize their variational free energy to adapt to the environmental uncertainty they encountered. At a specific time point in the simulation an out of distribution shock was introduced into the system. The findings provide evidence that organizations cannot rely on reductionism and must acknowledge the necessity of variability in order to produce a larger behavioural repertoire which will allow the organization to successfully utilize environmental uncertainty.